Difference between revisions of "Class meeting for 10-405 SGD for MF"
From Cohen Courses
Jump to navigationJump to searchLine 20: | Line 20: | ||
* Loss functions for matrix factorization that are appropriate for collaborative filtering | * Loss functions for matrix factorization that are appropriate for collaborative filtering | ||
* Algorithm and updates for SGD implementation of matrix factorization | * Algorithm and updates for SGD implementation of matrix factorization | ||
− | * | + | * DSGD algorithm - what is done in parallel and what is done sequentially |
− | * Definitions: stratum, interchangable steps | + | * Definitions: stratum (aka "diagonal"), interchangable steps |
Latest revision as of 12:39, 5 March 2018
This is one of the class meetings on the schedule for the course Machine Learning with Large Datasets 10-405 in Spring 2018.
Slides
Quiz
Papers Discussed
- Large-Scale Matrix Factorization with Distributed Stochastic Gradient Descent, Gemulla et al, KDD 2011.
Things to Remember
- Definition of matrix factorization
- Common applications of matrix factorization, and how they map into the MF problem
- Loss functions for matrix factorization that are appropriate for collaborative filtering
- Algorithm and updates for SGD implementation of matrix factorization
- DSGD algorithm - what is done in parallel and what is done sequentially
- Definitions: stratum (aka "diagonal"), interchangable steps